Scholarlink

<p>With the rapid advancement of wind power, energy storage, and hydrogen technologies, wind–hydrogen coupled energy systems are playing an increasingly vital role in enhancing renewable energy utilization and power supply reliability. However, multi-type maintainable components within such systems frequently encounter challenges during long-term operation, characterized by highly stochastic degradation, partially observable health states, and significant asymmetries between maintenance costs and failure consequences. These complexities render traditional strategies based on fixed intervals or static thresholds inadequate for achieving long-term economic optimality. Addressing these challenges, this paper models the maintenance optimization problem from a sequential decision-making perspective and proposes a system-level maintenance framework based on a Cost-Sensitive Health Index. Under conditions of incomplete information, the proposed method maps component degradation, failure risks, and economic consequences into a single, interpretable decision state variable. Building upon this, a system-level dynamic maintenance threshold policy is developed to coordinate the activities of multiple components under resource-constrained conditions. The corresponding decision problem is formulated as a long-term sequential optimization model, which is solved using reinforcement learning to learn adaptive dynamic thresholds. Finally, the Rye microgrid in Norway is used to demonstrate the proposed framework in stochastic operating environments. The results show that, compared with static rules, myopic decision-making, and policies that ignore system-level resource constraints, the proposed framework achieves improved long-term cost performance, balanced maintenance intensity, and reduced resource constraint violations.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Cost-sensitive health index-based maintenance optimization for smart wind-hydrogen systems via reinforcement learning

  • 期刊论文

  • 发表日期：

    2026年6月26日

查看全文
  • Hongyan Dui,
  • Rukun Wang,
  • Wanyun Xia,
  • Gang Liu,
  • Helin Li

摘要

With the rapid advancement of wind power, energy storage, and hydrogen technologies, wind–hydrogen coupled energy systems are playing an increasingly vital role in enhancing renewable energy utilization and power supply reliability. However, multi-type maintainable components within such systems frequently encounter challenges during long-term operation, characterized by highly stochastic degradation, partially observable health states, and significant asymmetries between maintenance costs and failure consequences. These complexities render traditional strategies based on fixed intervals or static thresholds inadequate for achieving long-term economic optimality. Addressing these challenges, this paper models the maintenance optimization problem from a sequential decision-making perspective and proposes a system-level maintenance framework based on a Cost-Sensitive Health Index. Under conditions of incomplete information, the proposed method maps component degradation, failure risks, and economic consequences into a single, interpretable decision state variable. Building upon this, a system-level dynamic maintenance threshold policy is developed to coordinate the activities of multiple components under resource-constrained conditions. The corresponding decision problem is formulated as a long-term sequential optimization model, which is solved using reinforcement learning to learn adaptive dynamic thresholds. Finally, the Rye microgrid in Norway is used to demonstrate the proposed framework in stochastic operating environments. The results show that, compared with static rules, myopic decision-making, and policies that ignore system-level resource constraints, the proposed framework achieves improved long-term cost performance, balanced maintenance intensity, and reduced resource constraint violations.